ARTICLE pubs.acs.org/EF
Characterizing Biomass Fast Pyrolysis Oils by 13C NMR and Chemometric Analysis Gary D. Strahan,† Charles A. Mullen,† and Akwasi A. Boateng* Eastern Regional Research Center, Agricultural Research Service, U.S. Department of Agriculture, 600 E. Mermaid Lane, Wyndmoor, Pennsylvania 19038, United States
bS Supporting Information ABSTRACT: Several biomass fast pyrolysis oils were characterized by 13C and DEPT (distortionless enhancement polarization transfer) NMR analysis to determine their chemical functional group compositions as related to their energy content. Pyrolysis oils were produced from a variety of feedstocks, including energy crops, woods, animal wastes, and oil seed presscakes, and were also compared with fossil fuels (gasoline and diesel). The 13C and DEPT NMR spectra of the biofuels and the two fossil fuel samples were measured, and each spectrum was divided into distinct regions according to the chemical shift ranges appropriate for various functional groups. The spectral intensities of each region were then quantified, and the carbon proton substitution numbers were determined. This approach provided helpful information on the chemical compositions of the oils, but it was found to oversimplify the information contained in the 13C NMR spectra, resulting in a substantial loss of important details. Hence, a chemometric method of analysis based on principle component analysis (PCA) was used to extract more of the useful information from the 13C NMR spectra. The intensities of the 13C NMR spectra from 15 pryolysis and two fossil fuel samples were binned in 2 ppm increments and subjected to unsupervised PCA. Comparison of the PCA scores plots with their respective loadings plots enabled a determination of the chemical shifts and, hence, the chemical functional groups that were most important in discriminating among the clusters. The clustering of the biomass samples based on PCA results were shown to track with their biological origin and their energy contents. Therefore, 13C NMR PCA analysis was shown to be a powerful and facile technique for classifying biomass fast pyrolysis oils.
’ INTRODUCTION Fast pyrolysis is one of the most promising methods for the conversion of biomass into a liquid product that can be ultimately used to produce advanced, drop-in biofuels. The pyrolysis oil (bio-oil) produced is an energy dense fuel intermediate that can be upgraded to hydrocarbons in the diesel, gasoline, and jet-fuel range fractions. Most research on the topic has focused on the conversion of lignocellulosic biomass such as wood, energy crops, and agricultural residues,1,2 though other biofeedstocks, including oil seed presscakes and animal wastes, have also been studied.3 6 In all cases, the as-produced fast pyrolysis oils are very complex mixtures with significant compositional variability typically influenced by the source material and exact pyrolysis process conditions. Pyrolysis oils are usually made up of oxygenated hydrocarbons of every functionality, including alcohols, ethers, ketones, aldehydes, esters, phenols, and carboxylic acids.1,2 For certain feedstocks, the mixtures can be further complicated by the presence of nitrogen containing compounds or hydrocarbons from plant resins or lipids.3,7 As such, the characterization of pyrolysis oils has been a challenging problem and usually requires the combined use of several analytical techniques, chromatographic (GC, HPLC, GPC) and spectroscopic (MS, IR) just to get a partial compositional picture of the mixtures.8 10 We have previously reported on the use of NMR, including 1H, 13C, and DEPT (distortionless enhancement polarization transfer) analysis, to characterize several fast pyrolysis oils, including those produced from woody materials, grasses, and legumes.11 The major advantage of NMR spectroscopy is the r 2011 American Chemical Society
ability to characterize the entire, nonfractionated pyrolysis oil. Since our previous report, other researchers have also carried out in-depth characterizations of pyrolysis oils using 1H and 13C NMR.12,13 We have expanded and improved upon our use and methods for analyzing 13C NMR to characterize fast pyrolysis oils by including 15 different biomass feedstocks and using principle component analysis (PCA) to compare them to each other and to the fossil derived fuels diesel and gasoline. The pyrolysis oils were derived from a wide array of feedstock including rye grass, switchgrass, alfalfa, pennycress presscake, pennycress defatted presscake, eel grass, oak wood, cow manure, barley straw, barley DDGS (dried distillers grains with solubles), camalina presscake, guayule plant, guayule bagasse, corn stover, and chicken litter. The production and base analyses of pyrolysis oil from each of these have been previously reported.3,7,14 16 As done previously, 13C and edited DEPT experiments were integrated and used to quantify carbon atoms in the pyrolysis oils based on attached protons and chemical environment. This study has focused primarily on the analysis of 13C and edited DEPT experiments, as the accessible information content in 13C NMR is much higher than with 1H NMR. 13C NMR also has the advantage of being less susceptible to chemical shift changes due to variations in pH and other environmental factors. This method not only provides Received: September 2, 2011 Revised: October 11, 2011 Published: October 12, 2011 5452
dx.doi.org/10.1021/ef2013166 | Energy Fuels 2011, 25, 5452–5461
Energy & Fuels
ARTICLE
Table 1. Percentage of Carbon Based on 13C NMR and DEPT analysis of Fast Pyrolysis Oils from Various Feedstocks as a Function of Chemical Shift (ppm) defatted ppm range,
carbon
functional groups
saturation
0 55, aliphatics
CH2 CH3
X2N CH2 R Ar CH2‑
55 95 , alcohols sugars RO CH2
Ar OCH3
95 165, aromatics, heteroaromatics 165 180, acids,
rye
barley
eel
cow
pennycress
pennycress
camelina
barley
oak
grass
straw
grass
manure
presscake
presscake
presscake
DDGS
gasoline
27.7
26.0
28.9
28.0
25.5
45.2
30.1
44.0
29.7
66.5
80.1
4.2
1.9
5.7
5.6
5.4
8.4
8.6
3.7
2.6
7.7
16.7
CH2
8.7
10.9
8.9
10.9
8.5
8.4
7.1
19.5
9.3
19.7
35.8
CH3
12.0
14.3
12.2
17.1
15.2
21.7
11.3
22.4
17.2
38.0
30.3
All
19.7
21.3
13.1
15.6
18.1
7.9
24.8
9.5
44.9
3.6
0.3
0.5
0.3 0
All CH1
CH0
diesel
CH1
10.1
9.8
9.2
6.6
8.9
1.6
17.0
2.9
28.3
0
CH2
3.4
6.0
4.7
3.3
3.2
1.5
0.9
0.4
17.2
3.1
0
CH3
3.8
3.0
1.0
0.3
2.5
0.3
0
3.7
0
0.0
0
All
44.3
43.9
50.7
32.9
40.1
40.4
28.1
39.0
19.3
28.9
CH0
32.5
24.7
24.8
17.9
20.7
15.7
11.4
5.3
14.0
8.9
19.03 5.8
CH1
17.3
19.1
25.5
14.5
18.5
24.0
16.7
33.0
5.2
19.9
13.0
All
7.6
6.6
4.6
16.9
12.9
5.3
12.7
7.5
5.0
0.6
0
All
0.7
2.2
2.8
6.7
3.4
1.2
4.3
0.0
1.0
0.38
0.62
esters, amides 180 215, ketones, aldehydes total
CH0
39.9
34.6
32.2
41.3
36.7
22.1
28.3
12.8
20.0
11.4
4.3
CH1
31.9
30.9
40.5
26.8
32.9
33.9
45.4
39.8
36.2
27.6
29.9
CH2
12.1
16.9
13.7
14.6
12.9
21.9
10.6
24.0
26.7
22.9
35.5
CH3
16.1
17.6
13.6
17.3
17.7
22.1
15.6
23.4
17.3
38.1
30.4
information about the different carbon chemical environments (based on chemical shift) but also establishes the proton substitution numbers of those carbon atoms. Although such exhaustive quantitative 13C analyses provide a very highly detailed picture of the chemical composition of each pyrolysis oil produced from the said feedstock, it is clumsy and impractical for comparing larger numbers of samples. In our prior analysis, these spectra were reduced to five chemical shift regions and analyzed as simple percentages of total 13C content.11 While this approach greatly simplifies the analysis, it also discards much of the rich information content contained in the data. A more rapid, efficient, and information-rich method for evaluating and comparing pyrolysis oils requires the development of chemometric models. In this paper, a model is developed and validated by comparison to the earlier, more standard approach. Once created, the model can then be used to test and examine pyrolysis oils from new and future potential feedstocks to gain an expedited understanding of the most significant components in a group of mixtures or of the available energy content, or to predict potential byproducts and uses. Similar analyses might be feasible by using 1H NMR alone, but 13C is intrinsically more informative owing to the greater chemical shift dispersion, separation of functional groups, and decreased environmental affect on chemical shifts. This approach may yield important information for determining what types of liquid fuel or chemicals might be best produced using pyrolysis oils themselves as a feedstock for downstream processes and/or applications or in the optimization of fuel blends,17 or for comparing production or purification methods.
’ METHODS Feedstocks. The pyrolysis oils in this study were produced from 15 different biomass feedstocks (those listed in Table 1). While they can be
grouped under various combinations of commodities, we choose to functionally group them into three categories: mostly lignocellulosic (woods, e.g. oak; grasses and crop residues, e.g. barley straw and rye grass), animal waste (i.e., cow manure and chicken litter), aquatic species (eel grass, zoestra), and high protein byproduct (i.e., pennycress and camelina presscakes and barley DDGS). Details on most feedstocks from which pyrolysis oils in this study were derived have been previously published.3,7,14 16 Gasoline and diesel samples were standard road vehicle fuels, with the former being a typical U.S. E-10 gasoline, a blend of petroleum distillate (∼90%) and ethanol (∼10%). Pyrolysis-Oil Production. For all samples, pyrolysis oil was produced by a fluidized-bed pyrolysis reactor over an inert silica sand medium at temperatures between 450 550 °C. Detailed descriptions of the pyrolysis system have been previously published.14,15 For each feedstock, the sample used for the NMR characterization was that fraction collected in the electrostatic precipitator (ESP), the last point in the condensation train, where most of the pyrolysis oil is collected and with the least moisture, typically 5 8 wt %. The higher heating values (HHV) of pyrolysis oils were determined using a bomb calorimeter (Leco AC3000). Nuclear Magnetic Resonance (NMR) Spectroscopy. Solutionstate (acetone-d6) NMR spectra were recorded at 9.4 T on a Varian Inova NMR spectrometer, using a 5 mm dual broad-band probe equipped with z-axis pulsed field gradients. All spectra were acquired at 45 °C, except for gasoline and diesel, which were acquired at 20 °C because of their volatility at this temperature. All 13C spectra, at 100 MHz, had a sweep-width of 30 000 Hz, were acquired using a 45° pulse angle and inverse-gating, and were referenced to the acetone methyl peak at 29.92 ppm. The t1 relaxation rate was measured for the whole guayule sample and the relaxation agent, chromiumacetylacetonate (Cr(acac)3), was added to reduce the maximum relaxation time to e5 s, without adversely affecting spectrum quality. An equal amount of relaxation agent was added to all other samples, except gasoline and diesel, resulting in maximum t1 relaxation times for the other samples that were typically 3 4 s. Reasonable signal-to-noise was generally 5453
dx.doi.org/10.1021/ef2013166 |Energy Fuels 2011, 25, 5452–5461
Energy & Fuels
ARTICLE
Figure 1. 1D-13C NMR spectra of the nine biomass pyrolysis-oils. For spectra of other pyrolysis oils see ref 11. achieved with ∼1024 transients, utilizing an acquisition time of 2.1 s, and an 8 12 s relaxation delay to provide adequate recovery of the signal for integration purposes. Fully edited DEPT experiments were run using the same parameters as were used for the 13C spectra but with standard flip angles of 45, 90, and 135°, and the coupling constant, JCH, was set to 135 Hz. In addition, a modified version of the edited DEPT spectrum was run in which the JCH-coupling was averaged over the values 135 175 Hz, at 5 Hz intervals, as has been done with HSQC (heteronuclear singlequantum correlation) spectra.19 This was done to evaluate and perhaps ameliorate the extent to which C H couple constants affect the final intensity in these complex mixtures. Both versions of the DEPT required approximately the same number of total scans and hence the same amount of time; both were mathematically manipulated to generate the CH, CH2, and CH3 subspectra. The data reported in this paper utilized the integrated values from the 1JCH-averaged edited DEPT experiment,
but a comparison of the integrated areas of the two methods suggests that, at least for these mixture samples, the differences were not significant within the errors of integration, although they may be more important with other types of samples. A standard sample mixture consisting of equal molar amounts [1:1:1 (mol)] of 3-ethylphenol, 4-hydroxy-4-methyl-2-pentanone, and 2-methyl-3-pentanone was analyzed using identical 13C and DEPT methods in order to estimate the relative effects on the signal intensities arising from the 1H NOE effect and the 1JCH-coupling in the DEPT experiment.18 Using the standard sample mixture as a guide, the intensity of each subspectrum was adjusted using experimentally determined correction factors to produce reasonable and consistent carbon integrations for carbons of each proton substitution number as a function of chemical shift. In this way, the individual spectra from the edited DEPT experiment were adjusted and compared directly to each other, providing an analysis of the approximate 5454
dx.doi.org/10.1021/ef2013166 |Energy Fuels 2011, 25, 5452–5461
Energy & Fuels
ARTICLE
scatter plots, loading plots, and contribution analyses.20,21 Specific chemical shift bins that were determined to be important (based on the loading plots and contribution analyses) were then compared with the DEPT analyses to determine the appropriate proton substitution number for those carbons and ascertain the likely functional groups. While wider bin sizes provided limited discrimination of the pyrolysis oils in the models, the 2 ppm bin size provided more detailed information about chemical moieties and may potentially be used to identify specific compounds. The PCA had difficulty fitting the data when spectral bins were used that were consistent with the five, very wide, chemical shift ranges utilized when analyzing the 13C and DEPT spectra in terms of percentages (as in Table 1).
’ RESULTS AND DISCUSSION
Figure 2. 1D-13C NMR spectra of the two fossil fuels, diesel and gasoline. percent CHn composition. The intensities of the peaks in the 13C and DEPT-45 spectra were normalized to each other using the entire spectrum between 0 and 95 ppm, excluding the region near the acetone solvent peak. By subtracting the integration values of the adjusted DEPT-45 spectra from those of the 13C spectra, an estimate of the percent of unprotonated carbons (CH0) was obtained for the 95 160 ppm region of the spectrum. As noted above, errors due to incomplete relaxation were reduced by the addition of a relaxation agent. This is important when quantifying chemical functionalities, but it is less critical when PCA is utilized to compare component mixtures. As in our prior work, the 13C and DEPT spectra were divided into five different spectral regions, roughly corresponding to different functional groups. The integrated area within each of the regions was then presented as a percentage of the whole to obtain a simple picture of the component mixture. Principle Component Analysis. The 1D-13C spectra were processed using Chenomx software (Alberta, Canada), and their chemical shifts were referenced to the methyl peak of the acetone solvent. The spectra were then binned in 2 ppm increments, excluding the acetone solvent regions (29 31 ppm and 204 211 ppm) and imported into Sweden) analysis software. The need to SimcaP-12 (Umetrics, Umea, exclude acetone solvent regions means that some potentially important overlapping pyrolysis oil signals could get removed from the analysis. While these gaps are expected to alter slightly the apparent compositions of the pyrolysis oils and alter their positions within their clusters based on principle component analysis, it is not expected that this will markedly affect the cluster to which each belongs. Because all useful solvents have potentially overlapping signals, we deemed this to be an allowable source of error. The spectra most affected by this gap in the methyl region are those of diesel, defatted pennycress, cow manure, camalina, eelgrass, and pennycress. The binned intensities were meancentered and normalized to the total spectral intensity, scaled by Pareto and modeled using unsupervised principle component analysis (PCA). The resulting models were visualized and analyzed using 2D score
Summary of 13C and DEPT analysis. The 13C and editedDEPT spectra were acquired for the nine additional pyrolysis oils plus the fossil fuels, gasoline and diesel (see Figures 1 and 2 for 13C spectra and the Supporting Information for full scale 13C and DEPT spectra). Their quantitative integrations are summarized in Table 1. Signals in the 0 55 ppm range represent aliphatic carbon atoms that are separated from oxygen atoms by at least two bonds, although those adjacent to nitrogen atoms can be found in this region. The pyrolysis oils derived from oak, rye grass, barley straw, eel grass, and cow manure all have 25 29% of their carbon in this category. The DEPT analysis reveals that the majority of these materials have about a 3:2:1 ratio of CH3:CH2:CH1 of nonoxygen adjacent aliphatic carbons. An exception was rye grass pyrolysis oil, which contained a smaller proportion of CH1, suggesting that fewer branch points are present in any hydrocarbon chains that are present in that material. Most of these pyrolysis oils, those derived from mostly lignocellulosic feedstocks, exhibited between 13 and 21% of its carbon resonating in the 55 95 ppm region, which represents carbon adjacent to heteroatoms (mostly oxygen) in alcohols, ethers, and anhydrous carbohydrates. In this region, there is generally a larger proportion of methine and methylene carbons (CH1 and CH2) compared to the proportion of methyl groups. Nearly all of the methyl groups in this region are probably methoxy groups on the 2 and 6 positions of phenolics rings in structures derived from lignin. These groups are in the highest proportion for oak and rye grass feedstocks and are lower for barley and eel grass, which could be indicative of the level of methoxy substitution and amount of lignin in the feedstocks and hence the pyrolysis oils. As anticipated, the fossil fuels have little or no resonance intensity in this region, as they contain very few carbons adjacent to heteroatoms. The aromatic region of the spectra (95 165 ppm) is where the most variability was seen in the pyrolysis oils from the mostly lignocellulosic feedstocks. This region includes aromatic carbon from both benzene rings and heteroaromatics (e.g., furans). The overall aromatic carbon content ranged from 33% for eel grass pyrolysis oil up to 50% for barley straw pyrolysis oil. Oak had the highest proportion of substituted aromatic carbon (CH0), which could be attributed to a high level of methoxy substitution on phenolic rings, because hardwoods tend to have higher methoxy content in the lignin than do softwoods or herbaceous grass species.22 This is also consistent with the observation of strong Ar OCH3 signals in the 56 ppm region. The extreme downfield end of the spectra represents carbonyl carbons with 165 180 ppm representing carboxylic acids, esters, and amides and 180 215 ppm representing ketones and aldehydes. Eel grass 5455
dx.doi.org/10.1021/ef2013166 |Energy Fuels 2011, 25, 5452–5461
Energy & Fuels pyrolysis oil was shown to have the highest concentration of these functional groups among the pyrolysis oils studied here. For pyrolysis oils derived from the presscakes of pennycress and camelina presscakes, the proportion of aliphatic carbon atoms is much greater than those in the lignocellulosic group, with aliphatics accounting for 45% of the carbon in these feedstocks. This is likely due to two factors: (1) the presence of residual triglycerides in the pyrolysis feedstock and (2) the higher nitrogen content in these feedstocks, as this region can also account for carbon adjacent to nitrogen atoms. The higher proportion of CH2 groups in the camelina presscake pyrolysis oil compared with the pennycress might suggest that the camelina presscake pyrolysis oil contains a greater proportion of long straight chains than that from pennycress. For pyrolysis oil produced from defatted pennycress presscake, which has had the entire amount of residual lipids removed (via hexane extraction), 30% of the carbons were still found in this aliphatic region, although the proportion of methyl (CH3) carbons was lower in this case. Pyrolysis oil from barley DDGS was similar to that from defatted pennycress with respect to the makeup of its spectra in the 0 55 ppm region. Each of the pyrolysis oils from the oil seed presscakes had CH X (72 75 ppm), as well as a limited variety of amides [196 183 ppm], and a relative lack of unsaturated carbons. For the barley DDGS sample, these resonances likely indicate the presence of amides derived from organonitrogen compounds, in turn, derived from protein-derived amines reacting with fatty acids resultant from triglycerides in the DDGS. Cluster B, consisting of pyrolysis oils from oak, switchgrass, rye grass, barley straw, corn stover, and cow manure, is distinguished by a relative lack of simple straightchain aliphatics ( CH3 and CH2 ) and increased significance of unsaturated >CH (arene, heteroaromatic, and/or vinylic), aliphatic heteroatoms, and some amides, esters, and ketones. The heteroaromatics and heteroaliphatics most likely arise from a mixture of sugars, condensed sugars, and alcohols. The compounds containing benzene rings (mostly phenolics) arise from products of lignin decomposition. Eel grass pyrolysis oil may fall into cluster B as well. In all of the scores plots, eel grass pyrolysis oil is the closest to the center, indicating that it is fairly average in its composition compared to the other pyrolysis oils in this study. This average nature of eel grass is similar to that found in a recent study that concluded that variations in protein content affect pyrolysis oil properties.23 An interesting observation that is revealed by an examination of the loadings plots is that the center of the plot has slightly more ketone contribution, indicating that eel grass pyrolysis oil also has more ketones compared to the other pyrolysis oils. This was also observed by direct analysis of the 13C spectra and is shown in Table 1. Cluster C consists of pyrolysates of whole guayule, guayule bagasse, and full flower alfalfa, and its significant distinguishing chemical signatures include unsaturated carbons such as nonheterocylic aromatics [116 133, 140 ppm], simple linear aliphatics ( CH2 and CH3) [28, 32 39 ppm], a limited number of heterosubstituted aliphatics ( CH2 X) [46 51 ppm], and a relative lack of heteroatoms and acids. Taken together, this suggests that the distinguishing chemical substances are aromatics substituted with mostly linear hydrocarbons (alkyls and alcohols). Cluster D contains the same pyrolysis oils that appeared to be most similar to the fossil fuels in model 1. Specifically, these are 5458
dx.doi.org/10.1021/ef2013166 |Energy Fuels 2011, 25, 5452–5461
Energy & Fuels
ARTICLE
Table 2. Higher Heating Values (HHV) of Pyrolysis Oils 13
pyrolysis oil HHV
C NMR PCA
a
clustera
pyrolysis oil feedstock
A
defatted pennycress presscake
29.1
A
barley DDGS
27.7
B
corn stover
24.3
B
eel grass
26.1
B
cow manure
28.8
B
rye grass
25.7
B
switchgrass
23.1
B
barley straw
26.3
B
oak
22.5
C
alfalfa stems
30.6
C
guayule (whole)
30.4
C
guayule bagasse
30.5
D
chicken litter
31.2
D
pennycress presscake
31.9
D
camelina presscake
32.0
E
gasoline (E10)
45.3
E
diesel
45.0
(MJ/kg, DB)
Figure 3.
pyrolysis oils from the camalina and pennycress presscakes and from chicken litter, and their significant distinguishing components are primarily long, simple, straight-chain aliphatics [12 17, 24 29 (mostly), 34 37 ppm], although there are also signatures arising from a limited set of aromatics [132 135 ppm] and an even more resonance-specific signature from carboxylic acids, esters, or amides [178 181 ppm]. These relatively tight ranges of resonances suggest that the underlying structures are, perhaps, more concentrated in a smaller range of structures, when compared with the compositions of other pyrolysis oils. This discussion is not meant to imply, however, that pyrolysis oils in cluster D have no acids, esters, amides, or unsaturated carbons but rather that any resonances arising from them are not as important in distinguishing them from other pyrolysis oils. As indicated, the third principle component only explains ∼10% of the variation in the samples, but it provides some additional interesting insights into their relative compositions. The scores plot of PC1 versus PC3 (Figure 6) depicts the clustered pyrolysis oils as being distributed in each of the four quadrants, more-orless along the 45° and 45° diagonals. On the basis of a comparison with the corresponding loadings plot, the upperright (NE) quadrant is associated with increasing heteroalkanes (especially secondary and tertiary structures, CH2 X and >CH X), but the lower-right (SE) is distinguished by heteroaromatic constituents, with increasing esters, amides, and/or carboxylic acids. In particular, cow manure pyrolysis oil is distinguished from the others by a comparatively larger and more varied concentration of carboxylic acids, amides, or esters [166 179 ppm]. Also noteworthy is that, in all previous scores plots, barley DDGS and defatted pennycress were in the same cluster, but in this plot, barley DDGS is in the extreme NE quadrant, while defatted pennycress is in the SE quadrant. This arises from the presence of heteroatoms in linear alkane structures, CH2 X, arising from alcohols or ethers [64 and 74 ppm] in the barley DDGS sample. The lower-left quadrant (SW), including pyrolysates of pennycress and camalina presscakes, is distinguished by increasingly significant aliphatic alkanes ( CH2 and CH3) [16 17, 24 29, 34 37 ppm], with a limited variety
of aromatic or alkenes [130 133 ppm] and only one important carboxylic acid resonance [178 ppm]. Pennycress presscake pyrolysis oil is barely in this SW quadrant and is actually closer to the upper-left (NW) quadrant. Thus, the SW quadrant primarily describes the salient features of camalina pyrolysis oil, rather than those of pennycress. Interestingly, chicken litter pyrolysis oil is no longer in the same group as these, as it has a less significant carboxylic acid composition, probably because of the lack of triglycerides (fat) in the feedstock. Instead, chicken litter pyrolysis oil appears in the upper-left (NW) quadrant, which is distinguished by aromatic hydrocarbons (112 133 ppm) and a diversity of aliphatic alkane structures (long and short, straight and branched) (16, 28, 30 54 ppm), as well as by a lesser amount of acids/esters/amides. Summation of 13C NMR PCA. Throughout all of the analyses, the pyrolysis oil members in each cluster remain relatively consistent. Clearly the pyrolysis oils of camalina, pennycress presscake, and chicken litter stand out in all of these analyses as being the richest in simple hydrocarbon alkanes, and the most like (though still significantly different from) the fossil fuels. The pyrolysis oils in cluster D (e.g., Figure 3) including those from oak, corn stover, switchgrass, rye grass, cow manure, etc. have the most typical biomass pyrolysis compositions. The significance of the utility of this 13C NMR PCA-based analysis method is demonstrated by examining the energy content of the various pyrolysis oils. The energy content of each of the pyrolysis oils was determined by means of bomb calorimetry, and these values are reported as higher heating values (HHV) in Table 2. The results show that the PCA clusters based on the binned 13C NMR intensities (as a function of chemical shifts) coincide with their energy contents, even though the HHVs were not used as input for the PCA analysis. Pyrolysis oils in cluster B (e.g., Figure 3), which contained the pyrolysis oils from the mostly lignocellulosic feedstocks, have the lowest HHV of all the samples studied, all between 22.5 and 26.3 MJ/kg (dry basis), with the exception of pyrolysis oil from cow manure which had an HHV of 28.8 MJ/kg. This may be due to the its higher concentration of amides, esters, and/or carboxylic acids, as indicated in the third principle component of model 2 (Figure 6). Additionally, the presence of additional high calorific value hydrocarbon-like compounds could be masked by the acetone solvents signals at 28 31 ppm (see the Methods section). Cluster A, consisting of the pyrolysis oils from defatted pennycress presscake and barley DDGS, had HHVs between 27.7 and 29.1 MJ/kg. The HHVs for cluster C, consisting of the pyrolysis oils from guayule and alfalfa stems, are the next highest, with energy contents between 30.4 and 31.2 MJ/kg. Of all of the biomass fast pyrolysis oils considered in this study, those in cluster D had the highest values. These are the pyrolysis oils from the oil seed presscakes and chicken litter, and they have HHVs between 31.0 and 31.2 MJ/kg. The clustering of the pyrolysis oils into these groups demonstrates the relationship between chemical structures and physical properties and how using the 13C NMR PCA analysis provides an efficient method for classifying biomass pyrolysis oils.
’ CONCLUSIONS We have continued our in-depth NMR analysis of in-tact pyrolysis oils by including those from additional feedstocks, quantifying the chemical components on the basis of functional groups and comparing them with the fossil fuels diesel and 5459
dx.doi.org/10.1021/ef2013166 |Energy Fuels 2011, 25, 5452–5461
Energy & Fuels gasoline. A new reference standard mixture was used as it better suits these samples, and a JCH-averaged edited DEPT experiment was attempted to reduce the effects of unequal C H coupling, although we did not observe a significant benefit by this method for the types of samples studied here. The methodology was expanded by the use of unsupervised principle component analysis (PCA) of the 13C NMR spectra, which was validated by comparison to the results obtained using our prior methods to ensure an accurate understanding of the sample analyses. As noted in the Methods section, PCA had difficulty fitting the data when the binned intensities were the same as the five spectral regions used in the percentage-based chemical functional group analysis of the 13C and DEPT spectra. This indicates that reducing the 13C spectra into only five regions diminishes the information content by too great an amount. However, when smaller bin sizes were used, PCA provided excellent discrimination of the pyrolysis oils. It quickly discriminated among pyrolysis oil compositions based on chemical functional groups without substantial interpretative effort, and the resulting clusters correlated well with both biological origins of the oils and their enthalpy of combustion. This demonstrates that, by using PCA, pyrolysis oil samples can be quickly analyzed, compared to, and tested against an existing model to evaluate salient features and probable energy contents, without the need of purification, extraction, separation, or other special handling. This method is considerably faster and more unbiased than previous methods, has the promise of very rapidly and accurately providing significant information for pyrolysis oil chemical compositions, and may be applied to other similarly complex biomixtures. Future studies could include a statistical evaluation of the energy content of the pyrolysis oils and even the ease of purification, production, or growth. The use of supervised techniques might even enable classifications on the basis of particular chemical signatures.
’ ASSOCIATED CONTENT
bS
Supporting Information. PCA loading plots, an example interpretation of those plots, and full 13C and DEPT NMR spectra. This information is available free of charge via the Internet at http://pubs.acs.org/.
’ AUTHOR INFORMATION Corresponding Author
*Phone: 215-233-6493. Fax: 215-233-6559. E-mail: akwasi.
[email protected]. Author Contributions †
These authors contributed equally to this work.
’ DISCLOSURE Disclaimer: Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the U.S. Department of Agriculture. USDA is an equal opportunity provider and employer. ’ ACKNOWLEDGMENT The authors would like to thank David Chang (Chenomx) and Tamara Byrne (Umetrics) for helpful discussions.
ARTICLE
’ REFERENCES (1) Mohan, D.; Pittman, C. U., Jr.; Steele, P. H. Pyrolysis of Wood/ Biomass for Bio-Oil: A Critical Review. Energy Fuels 2006, 20, 848–889. (2) Huber, G. W.; Iborra, S.; Corma, A. Synthesis of Transportation Fuels from Biomass: Chemistry, Catalysts, and Engineering. Chem. Rev. 2006, 106, 4044–4098. (3) Boateng, A. A; Mullen, C. A.; Goldberg, N. M. Producing Stable Pyrolysis Liquids from the Oil-Seed Presscakes of Mustard Family Plants: Pennycress (Thlaspi arvense L.) and Camelina (Camelina sativa). Energy Fuels 2010, 24, 6624–6632. (4) Sricharoenchaikul, V.; Atong, D. Thermal Decomposition Study on Jatropha curcas L. Waste Using TGA and Fixed Bed Reactor. J. Anal. Appl. Pyrolysis 2009, 85, 155–162. (5) Schnitzer, M. I.; Monreal, C. M.; Facey, G. A.; Fransham, P. B. The Conversion of Chicken Manure to Bio-Oil by Fast Pyrolysis I. Analyses of Chicken Manure, Bio-Oils and Char by 13C and 1H NMR and FTIR Spectrophotometry. J. Environ. Sci. Health, Part B 2007, 42, 71–77. (6) Mante, O. D.; Agblevor, F. A. Influence of Pine Wood Shavings on the Pyrolysis of Poultry Litter. Waste Manage. 2010, 30, 2537–2547. (7) Boateng, A. A.; Mullen, C. A.; Goldberg, N.; Hicks, K. B.; McMahan, C. M.; Whalen, M. C.; Cornish, K. Energy-Dense Liquid Fuel Intermediates by Pyrolysis of Guayule (Parthenium argentatum) Shrub and Bagasse. Fuel 2009, 88, 2207–2215. (8) Oasmaa, A.; Kuoppala, E.; Solantausta, Y. Fast Pyrolysis of Forestry Residue. 2. Physicochemical Composition of Product Liquid. Energy Fuels 2003, 17, 433–443. (9) Mullen, C. A.; Boateng, A. A. Chemical Composition of Bio-Oils Produced by Fast Pyrolysis of Two Energy Crops. Energy Fuels 2008, 22, 2104–2109. (10) Garcia-Perez, M.; Chaala, A.; Pakdel, H.; Kretchmer, D.; Roy, C. Characterization of Bio-Oils in Chemical Families. Biomass Bioenergy 2007, 31, 222–242. (11) Mullen, C. A.; Strahan, G. D.; Boateng, A. A. Characterization of Various Fast-Pyrolysis Bio-Oils by NMR Spectroscopy. Energy Fuels 2009, 23, 2707–2718. (12) DeSisto, W. J.; Hill, N.; Deis, S. H.; Mukkamala, S.; Joseph, J.; Baker, C.; Ong, T.-H.; Stemmler, E. A.; Wheller, M. C.; Frederick, B. G.; Van Heiningen, A. Fast Pyrolysis of Pine Sawdust in a Fluidized-Bed Reactor. Energy Fuels 2010, 24, 2642–2651. (13) Joseph, J.; Baker, C.; Mukkamala, S.; Beos, S. H.; Wheeler, M. C.; Desisto, W. J.; Jensen, B. L.; Frederick, B. G. Chemical Shifts and Lifetimes for Nuclear Magnetic Resonance (NMR) Analysis of Biofuels. Energy Fuels 2010, 24, 5153–5162. (14) Boateng, A. A.; Daugaard, D. E.; Goldberg, N. M.; Hicks, K. B. Bench-Scale Fluidized-Bed Pyrolysis of Switchgrass for Bio-Oil Production. Ind. Eng. Chem. Res. 2007, 46, 1891–1897. (15) Boateng, A. A.; Mullen, C. A.; Goldberg, N.; Hicks, K. B.; Jung, H. G.; Lamb, J. F. S. Production of Bio-Oil from Alfalfa Stems by Fluidized-Bed Fast Pyrolysis. Ind. Eng. Chem. Res. 2008, 47, 4115–4122. (16) Mullen, C. A.; Boateng, A. A.; Goldberg, N. M.; Hicks, K. B.; Moreau, R. A. Analysis and Comparison of Bio-Oil Produced by Fast Pyrolysis from Three Barley Biomass/Byproduct Streams. Energy Fuels 2010, 24, 699–706. (17) Monteiro, M. R.; Ambrozin, A. R. P.; Liao, L. M.; Boffo, E. F.; Pereira-Filho, E. R.; Ferreira, A. G. 1H NMR and Multivariate Calibration for the Prediction of Biodiesel Concentration in Diesel Blends. 2009, 86, 581-585. (18) Henderson, T. J. Sensitivity-Enhanced Quantitative 13C NMR Spectroscopy via Cancellation of 1JCH Dependence in DEPT Polarization Transfers. J. Am. Chem. Soc. 2004, 126, 3682–3683. (19) Heikkinen, S.; Toikka, M. M.; Karhunen, P. T.; Kilpel€ainen, I. A. Quantitative 2D HSQC (Q-HSQC) via Suppression of J-Dependence of Polarization Transfer in NMR Spectroscopy: Application to Wood Lignin. J. Am. Chem. Soc. 2003, 123, 4362–436. (20) Webb-Robertson, B. J.; Lowry, D. F.; Jarman, K. H.; Harbo, S. J.; Meng, Q. R.; Fuciarelli, A. F.; Pounds, J. G.; Lee, K. M. A Study of 5460
dx.doi.org/10.1021/ef2013166 |Energy Fuels 2011, 25, 5452–5461
Energy & Fuels
ARTICLE
Spectral Integration and Normalization in NMR-Based Metabonomic Analyses. J. Pharm. Biomed. Anal. 2005, 39, 830–836. (21) Craig, A.; Clorarec, O.; Holmes, E.; Nicolson, J. K.; Lindon, J. C. Scaling and Normalization Effects in NMR Spectroscopic Metabonomic Data Sets. Anal. Chem. 2008, 78, 2262–2267. (22) Rodigues, J.; Meier, D.; Faix, O.; Pereira, H. Determination of Tree to Tree Variation in Syringyl/Guaiacyl Ratio of Eucalyptus globulus Wood Lignin by Anlayitcal Pyrolysis. J. Anal. Appl. Pyrolysis 1999, 48, 121–128. (23) Mullen, C. A.; Boateng, A. A. Production and Analysis of Fast Pyrolysis Oils from Proteinaceous Biomass. Bioenergy Res. Published Online First June 7, 2011, DOI: 10.1007/s12155-011-9130-x.
5461
dx.doi.org/10.1021/ef2013166 |Energy Fuels 2011, 25, 5452–5461